A Survey of Scheduling Algorithms in Cloud
Computing
Njoud AlMansour
Department of Computer Science, College of Computer Science
and Information Technology, Imam Abdulrahman Bin Faisal
University, P.O. Box 1982, Dammam 31441, Saudi Arabia
2190500057@iau.edu.sa
Abstract— In cloud computing, resources are provided as a
public utility and the user can lease and release those resources via
the Internet by an on-demand fashion. Since cloud resources are
meant to be utilized properly, the role of scheduling algorithms is
vital to ensure an appropriate resource is available to every
request. In this paper, major cloud computing scheduling
algorithms are reviewed, and discussion is made accordingly. This
paper also compares cloud computing scheduling from the
perspective of makespan, load balancing, CPU utilization,
deadline, response time, and allocation cost. In addition, the paper
proposes an abstract model to integrate desirable features the of
algorithm suitable to cloud environment. Future research
opportunities are highlighted and the end of the paper.
Keywords— Cloud Computing, Scheduling Algorithms, Model,
Survey
I. INTRODUCTION
Cloud computing concept goes back to the 1960’s when
John McCarthy [1] described computation as “computation may
some-day be organized as public utility”. Since then efforts have
been made and Amazon helped the emerging of cloud
computing development by launching Amazon web services as
a utility in 2006 [1]. Before this era, users relied on grid
computing which was seen as an infrastructure for providing raw
computing power like in compute grids or storage of date like in
data grids [2]. A new computing paradigm was found the
following years which is cloud computing, in this paradigm the
resources are provided as a public utility, in a pool, and the user
can lease and release those resources via the Internet by an on-
demand fashion [3]. The cloud computing has a service delivery
model shown Figure 1 in which consists of three layers: (i)
Infrastructure as a Service (IaaS) which has the basic hardware
to access the service and a collection of those hardware is what
forms the datacenter, (ii) Platform as a Service (PaaS) which has
the requires operating systems and database management
systems, and (iii) Software as a Service (SaaS) which has the
applications that are provided to the user in order to use cloud
services [4], [5].
978-1-5386-8125-1/19/$31.00 ©2019 IEEE
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on March 12,2021 at 07:01:53 UTC from IEEE Xplore. Restrictions apply.
Nasro Min Allah
Department of Computer Science, College of Computer Science
and Information Technology, Imam Abdulrahman Bin Faisal
University, P.O. Box 1982, Dammam 31441, Saudi Arabia
nabdullatief@iau.edu.sa
Fig. 1. Cloud Delivery Model
Since the resources in cloud computing is a public utility
many users can place a request for those resources at the same
time and this calls a need for scheduling to ensure the service it
available for every user which is a challenge given the fact that
there are many scheduling algorithms to consider and each one
of those algorithms has its own factors to keep in calculation
whether it is a cost matrix or QoS requirement [1]. Scheduling a
task in a system is a tedious procedure and to accomplish it, you
need to define few parameters and attributes that are related to
the system in which we are scheduling the task or related to tasks
we are scheduling. For scheduling in the cloud, we choose the
most suitable resources for the execution of a given task in way
that provide the minimal time required for the completion of a
task [6]. Types of scheduling a task can be divided into two
major types static scheduling where it has complete knowledge
about the structure of the tasks and the mapping of resources
beforehand and it estimates the execution time of the tasks and
the other type is dynamic scheduling which does not only
depend on the submitted tasks to the cloud environment but also
on the system state and the computer machines [6]. In general,
an efficient and optimized scheduling algorithm in cloud
computing need to consider the following issues like cost, time
and service level agreement (SLA) parameters to be followed to
as intended by the users [7]. There are also some parameters to
consider when choosing a scheduling algorithm like makespan,
load balancing, processor utilization, deadline, execution time,
completion time, and scalability [6]. In this paper we survey
scheduling algorithms used in cloud computing in various
settings referencing the applicable scheduling algorithms
parameters for each algorithm and at the end providing a rich
listing of cloud scheduling algorithms. At the end of the paper
an abstract model to facilitate the choice of the suitable
algorithms is proposed.
We divide the remaining part of this paper into 5 sections.
Findings are reported in Section II, while state of the art results
are provided in Section III. The detailed of chooser model are
described in Section IV while conclusion and future work is
highlighted in Section V.
II. FINDINGS
In this section, the results of the literature search and
selection is shows in two forms which are search execution
results and classifies results.
A. Search execution results
The total number of papers found were 32 publications.
After applying the inclusion, exclusion criteria 29 publications
remained. Finally, after applying quality assessment 27
publications remained.
B. Classified results
Here those 27 publications are classified depending on goal
and the outcome of those papers. The classification percentages
are shown in Table II below. The percentages show how many
papers are under that class.
TABLE I. RESULTS CLASSIFICATION BASED ON GOAL AND OUTCOME
Goal and outcome Percentage
Model, application, product, and/or service 47%
Review and survey 32%
Challenges and risks 21%
III. STATE OF ART
This section of the paper presents related work done in the
field of tasks scheduling algorithms in cloud computing. It is
divided into two sections: (i) Reviews and Surveys, that
mentions some of the reviews and survey done for scheduling
algorithms in the cloud including a review those algorithms
proposed by different authors, and (ii) Tabular Comparison of
Reviewed Algorithms, which discusses those algorithms that
researchers have introduced for the cloud scheduling in regard
to parameters like makespan, load balancing, CPU utilization,
deadline, response time, and allocation cost.
A. Reviews and Surveys
This section includes the reviews and the surveys done in the
field of cloud computing which acted as a building stone for
improving existing ones. The authors in [8] a review on
scheduling algorithms of virtual machines in cloud computing
where in mentioned the importance of improving existing
conventional scheduling algorithms like First Come First Served
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on March 12,2021 at 07:01:53 UTC from IEEE Xplore. Restrictions apply.
(FCFS) since they are not suitable for the dynamic nature of the
cloud in which the pool of resources is allocated dynamically.
They have done a narrative listing of the previous work done on
improving the traditional methods whether it is a model,
framework, or an algorithm [8]. At the end they have extracted
some information from that review which was mainly discussing
how conventional algorithms is costly when applied as a virtual
machine scheduling algorithms in the cloud compared to those
improved algorithms with the intelligence component like
Genetic Algorithms (GA) and Particle Swarm Optimization
(PSO) [8]. Many factors should be considered when choosing a
scheduling algorithm and it mainly dependent on the nature of
the task. In [9] the have considered the factors of allocations cost
which is the cost that a task requires when it is scheduled on the
cloud like memory cost for example the other factor is the load
balancing factor which is balancing the tasks on the running
nodes to distribute the load equally. They compared the
traditional FCFS algorithm with some of the optimized and
enhanced ones like PSO and Activity Based Costing (ABC) in
the CloudSim simulator and the conclusion was that some
algorithms performed well in regard to the load balancing factor
while failed to reduce the allocation cost factor but mostly the
traditional method FCFS required more allocation cost than the
algorithms it was compared to [9]. The authors in [10]
investigated the scheduling algorithms with regard to Quality of
Service (QoS) since it is an importance parameter for the user
satisfaction, where they did a flow chart explaining each
algorithms discussed in the paper to facilitate the extraction of
the advantages and disadvantages and hence draw a conclusion
about the studies algorithms. A survey on many task scheduling
algorithms is done in [11] in which they did a comprehensive
review on most of the existing algorithms by dividing them into:
economic, heuristics, and priority categories and then further
dividing those categories into subsection to facilitate the
classification of those algorithms by the reader. At the end they
produced a paper that can act as a reference for a quick overview
or a background search of a certain algorithm [11]. Another
survey is done by authors in [12] where they focused their study
on static algorithms in which they gave a brief definition of each
algorithm and then constructed a performance matrix to evaluate
and compare those algorithms. They focused their matrix on
make span, time complexity, resource allocation, QoS, average
waiting time, and average response time [12]. At the end they
provided future work opportunities based on the evaluation they
have done on the performance matrix [12]. A review on some
scheduling algorithms is done in [6] in which the author
provided an overview on FCFS, Shortest Job First (SJF), Round-
Robin (RR), and Priority scheduling algorithms. At the end the
author of [6] specified the parameters and the objectives for each
one of those algorithms which can act as a reference for
proposing new algorithms with a certain objective or enhancing
existing ones to achieve more objectives. A review on
scheduling algorithms based on meta-heuristics, which is the
modification of heuristics algorithms to achieve better
performance, method was done in [7] in which they divided the
meta-heuristics algorithms to the following classes: Bio-based
like GA, Swarm intelligence like PSO, Bat algorithms, and Cat
Swarm Optimization. This kind of division is helpful as the
heuristic algorithms act as the foundation of many successful
new introduced algorithms used for the cloud since scheduling
a task in the cloud is considered an Np- complete problem and
the best way to deal with it is to improve on heuristics hence
having this division can facilitate the way to researchers to
introduce more algorithms for the cloud environment. Author
summarized the issues found in [7] providing many
opportunities for R&D activities one of them is security
aware scheduling which is a major issue since cloud computing
relay heavily on conduction all its procedures virtually hence a
strong algorithm based on heuristics would act as a practical
solution for takes requiring security one of the algorithms to
consider is a discrete PSO [7]. The author in [5] discusses how
necessary it is to use optimized algorithms for cloud computing
rather than the traditional algorithms which are not adapted to
the dynamic nature of the cloud. For the main part of the paper
the author did a narrative listing if the previous work done in the
field and compared them using graphs from the reference studies
[5]. A scheduling algorithm based on GA called Tournament
Selection Genetic Algorithm (TS-GA) in [13] which has the
principle of when a good selection is found instead of removing
from the population it is instead introduced to the population
while another selection process starts. The TS-GA was
compared to RR and GA and registered 32% and 60% more
utilization and 7% and 8% less execution costs, respectively
[13]. Another algorithm based on GA was proposed in [14]
called Improved Genetic Algorithm Task Scheduling (IGATS)
which introduces the concept of load priority which was
compared to GA in terms of response time and execution time
and performed better on both parameters. Credit based
scheduling algorithm is proposed in [15] which considers two
parameters, task length and user priority. Since scheduling a task
based on task length only may cause starvation and scheduling
based on user priority only can decrease the utilization of
resources. The authors in [15] compared the proposed algorithm
with those focusing on either task length or user priority in terms
of makespan and found that the proposed algorithm performed
better. In [16] called Master-Service Uniform Multi-Round
(MSUMR) which adopt the conventional master-service model
along with Uniform Multi-Round (UMR) with the included
restriction of checking the network bandwidth. The performance
of MSUMR in terms of completion time was compared to UMR,
Multi-Installment (MI), eXtended Multi-Installment (XMI), and
One-Batch which are conventional single-round algorithms and
MSUMR showed better performance[16]. In [17] the authors
introduces a new technique called GreenSched for task
allocation and providing resources which is described as an
intelligent method since it is an energy aware. Having an energy
aware technique that uses CPN-based Green Cloud Scheduling
(CGCS) and Forward-only CPN-based scheduling (FCS) in
cloud is very useful since energy consumption is usually
considered to be a growing problem [17]. CGCS helps locating
the nodes with the least energy consumption where FCS
identifies the best nodes to use for scheduling [17]. This
techniques will still provide QoS measures by fulfilling the
deadline and budgets parameters which are specified by the user
[17]. They performed some experiments on this technique to
prove its application in which authors tested the speed of
deadline fulfillment in this techniques by using CGCS, FCS, and
SPECweb 2009 [17]. The techniques using CGCS and FCS
showed better results that those of SPECweb 2009 [17]. Another
paper in which the authors examined the deadline fulfillment is
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on March 12,2021 at 07:01:53 UTC from IEEE Xplore. Restrictions apply.
[18] where they proposed a solution to solve the problem of
scheduling tasks with deadline and cost constrains on both
public and private cloud. They discussed hybrid cloud which
helps in determining the where to places the available workload
whether it is on a public or private cloud [18]. They introduced
a model consisting of two components, public cloud scheduling
and hybrid cloud scheduling [18]. In the hybrid the algorithms
FCFS and Earliest Deadline First (EDF) are used and there are
two parameters extracted which assess whether to run this task
load on the public cloud or not and they the unfeasibility and the
cheapness [18]. In terms of deadline and allocation cost they
noticed after performing their experiments that EDF in both
feasible and cheapest settings had the highest deadline meeting
percentages with increasing the private cloud CPU and lowest
cost over application [18]. In [19] the authors introduced a new
algorithm based on Ant Colony Optimization (ACO) algorithm
called Load Balancing Ant Colony Optimization (LBACO)
algorithm. This algorithms will help in dynamically allocating
resources for tasks in the cloud [19]. Their experiments
compared their proposed algorithm against ACO and FCFS in
terms of load balancing and makespan [19]. The results show
that LBACO performed better in both parameters [19]. The
authors in [20] proposed a new meta-heuristics cost-effective
GA which is designed to meet the deadlines while keeping the
cost minimal. The algorithm is called Cost Effective Genetic
Algorithm (CEGA) [20]. In their experiments they compared
their proposed algorithm against IaaS Cloud Partial Critical Path
(IC-PCP), Robust-Cost-Time (RCT), Robust-Time-Cost (RTC),
and PSO algorithms in terms of makespan and allocation cost
[20]. The proposed algorithms perform better than any other
algorithm in makespan parameter whereas in allocation cost
parameter it performed better than RTC, RCT, and PSO [20]. In
this paper [21] the authors discussed the Multi-Population
Genetic Algorithm (MPGA) where the load balancing parameter
is take into consideration. The reason they chose the MPGA
instead of the GA approach is to avoid the convergence
prematurely and they found at the end of the paper that this
approach is more suitable for a large number of tasks [21]. Their
proposed approach was compared against Time and Cost
constrains Genetic Algorithm (TCGA) and Simulated
Annealing Genetic Algorithm (SAGA) in term of load
balancing, allocation cost, and response time [21]. MPGA
showed a better performance in both parameters [21]. In [22] the
authors proposed a new algorithms based on combining ACO
and PSO called Ant Colony Optimization with Particle Swarm
(ACOPS). ACOPS uses the previously saved information for the
prediction and adaptation of a new task sets [22]. This algorithm
was compared against ACO, FCFS, RR, GA, Aimulated
Annealing (SA), Predication mode based Routing Algorithm
based on ACO (PRACO), and Ant Colony System Tree Growth
(ACS-TG) in terms of load balancing [22]. The experiment
result showed that ACOPS better than other algorithms [22]. In
[23] the authors proposed algorithm based on PSO with
consideration to load balancing parameters. For this adaptation
the authors assumed that the tasks are non-preemptive and
independent [23]. The proposed algorithm was compared to RR
and improved PSO in terms of makespan, response time, and
CPU utilization [23]. In all comparison the proposed algorithm
showed better results [23].
B. Tabular View of Reviewed Algorithms where the simulation is done in a server and then the
The makespan, load balancing, CPU utilization, deadline, appropriate algorithm can be chosen based on that.
response time, and allocation cost of the algorithms and
techniques reviewed above from [13]–[23] are highlighted in IV. WORKING OF CHOOSER MODEL
Table III. Based on literature review done in this paper it is clear that
the choice of the scheduling algorithm is a difficult yet an
important one. Hence an abstract model shown in Figure 2 is
proposed to help in solving this problem. This model is
appropriate to be used in companies that execute many of their
TABLE II. TABULAR COMPARISON OF REVIEWED ALGORITHMS AND transactions on the cloud since it can be customized to fit the
TECHNIQUES needs of the users.

A. Model’s Components
Utilization

Allocation
Algorithm

Makespan

Balancing

Response
Deadline In this section the components of the model in Figure 2
which are User, Check Server, Chooser Server, and Cloud are

Time
Load

CPU

Cost
explained. The algorithms in Figure 2 and Figure 4 use some
notations and they are described as follows:
TS-GA [13] X High X Low Low
• U is a set of set users parameters received from the
IGATS [14] X X X X Low X
checker server
Credit Based Low X X X X X • S is a set of task sets received from the checker server
[15]
MSUMR [16] X X X X Low X • x is one task set from the S set of task sets
GreenSched X X X High X X • t is one task from x task set
[17]
• B in one bundle which is a set of many t
EDF X X X High X Low
(unfeasible • A is one attribute used to form B
and cheapest)
[18] • G is one chosen algorithm
LBACO [19] Low High X X X X
CEGA [20] Low X X X X Lower
than
(RTC,
RCT,
PSO)
MPGA [21] X High X X Low Low
ACOPS [22] X High X X X X
PSO with Low X High X Low X
Load
Balancing Fig. 2. An Overview of Chooser Model
[23]
1) User: The first component in this model is the user who
acts as the supplier of the task set to be executed in the cloud.
It can be seen that in each paper the authors compare their
The user is associated with two steps of the model’s execution
proposed algorithms to some of the conventional ones and each
author chose different parameters and criteria to consider when which are, supplying and outputting. This model can support
evaluating the proposed algorithm. This make the job of many users simultaneously.
choosing the needed algorithm to achieve a task difficult since 2) Checker Server: The second component of the model is
there are no common grounds which they can follow to the checker system who is associated with six steps of the
compare the available algorithms. To compare two algorithms, model’s execution which are, supplying, checking, reserving,
you have to know the algorithm steps clearly, which some answering, responding, and outputting. This server will set a
authors do not state in their papers and you also must execute it timer and collect many users’ entries during that time and then
in a simulator where you create an environment to run and perform its main role. The checker’s role is to check two things,
compare algorithms this can take a substantial time and require the availability of resources in the cloud to execute a task set
a thorough knowledge of the simulator tool. Hence an abstract
and checking for appropriate scheduling algorithm by asking
model, called chooser model, to help in choosing the
appropriate algorithm is proposed in section 5 of this paper the chooser server. When checking the resources in the cloud if
the resources are available and the appropriate scheduling

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on March 12,2021 at 07:01:53 UTC from IEEE Xplore. Restrictions apply.
algorithms is chosen then those recourses are reserved. This
server in its simplest form can be seen as the communication
layer between the user and the cloud and can enhance the
running time of many existing algorithms since it reserves the
resources beforehand. In Figure 3 below is an algorithm
showing the working of this server in an abstract fashion.
Fig. 3. Pseudocode -code of Checker Server Algorithm
3) Chooser Server: The third component of the model is the
chooser server which is associated with three steps of the
model’s execution which are checking, answering, and sending.
This server does not communicate directly with the user. The
chooser server role is to choose the appropriate scheduling
algorithm to be used for the task set in the cloud. Many users
can use this model hence many tasks sets may arrive at the same
time hence another role of this component is to organize the
tasks that are independent from different tasks sets by a
common attribute and group them in bundles and assign them
to the best suited scheduling algorithm based on the supplied
parameters by the checker server. The choice of the algorithms
is done by applying a Machine Learning (ML) model like
Support Vector Machine (SVM) [24] which can compare many
algorithms against each other and choose the one with highest
match rate in term of meeting the need of the users. Also, this
server communicates directly with the cloud in which it sends
the task set and the chosen algorithm. In Figure 3 below is an
algorithm showing the working of this server in an abstract
fashion.
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on March 12,2021 at 07:01:53 UTC from IEEE Xplore. Restrictions apply.
Fig. 4. Pseudocode -code of Chooser Server Algorithm
4) Cloud Server
B. Model’s Execution
In this section the execution of the model given in Figure 2
is explained. The execution of the model has many cycles and
the order in which those steps are listed does not necessary
follow the in sequence.
1) Supplying: The supplying step is where the user supplies
the task set to the check server and with that task set the user
can specify other parameters like QoS, cost, response time, etc.
2) Chekcing: In this step the checker server checks the
chooser server to see if the appropriate scheduling algorithm
exists. In this step the checker server supplies the parameters
from the user to the chooser server along with the task set.
3) Reserving: In this step the checker server checks the
availability of the needed resources to be supplied to the task
set execution after choosing the appropriate scheduling
algorithm, if there exist the necessary recourses for the task set
then those recourses are reserved. This can be done by placing
the task set in the queue of those resources. The address of the
checker server is saved in this step for the return of the output.
4) Answering: In this step the chooser server replays to the
checker server if the appropriate algorithm is found after that
the checker server can reserve the appropriate resources in the
cloud for execution.
5) Sending: This step is where the chooser server sends the
task sets and the chosen algorithms to be executed in the cloud
within the reserved resources which were reserved in the
reserving step.
6) Responding: This step is executed after the task sets
finish executing in the cloud then the results will be sent back
to the checker server using the address saved in the reserving
step.
 7) Outputting: The checker server in this step simply sends
the output from the cloud to the user.
V. CONCLUSION AND FUTURE WORK
Various cloud computing algorithms were discussed, and
classification was made from the perspective of makespan, load
balancing, CPU utilization, deadline, response time, and
allocation cost for such algorithm, based on the analysis, an
overview of an abstract model is presented that is customizable
as per transaction needs of companies. As a future work, it will
be interesting to extend the abstract model by adding artificial
intelligence, providing the provision of priorities important
tasks, enabling user to select appropriate algorithm and
maintaining secure communication during transactions.
REFERENCES
[1] K. I. Suthakar and M. K. K. Devi, “Resource Scheduling for Big Data on
Cloud,” pp. 185–205.
[2] F. Oesterle, S. Ostermann, R. Prodan, and G. J. Mayr, “Experiences with
distributed computing for meteorological applications: Grid computing
and cloud computing,” Geosci. Model Dev., vol. 8, no. 7, pp. 2067–2078,
2015.
[3] Q. Zhang, L. Cheng, and R. Boutaba, “Cloud computing: State-of-the-art
and research challenges,” J. Internet Serv. Appl., vol. 1, no. 1, pp. 7–18,
2010.
[4] K. M. Uma Maheswari and S. Govindarajan, “A survey of various cloud
scheduling algorithms,” J. Adv. Res. Dyn. Control Syst., vol. 9, no. 7, pp.
1–8, 2017.
[5] G. T. Hicham, “Optimization of Task Scheduling Algorithms for Cloud
Computing : A Review Optimization of Task Scheduling Algorithms for
Cloud Computing : A Review,” no. October, pp. 664–672, 2017.
[6] E. Kumari, “a Review on Task Scheduling Algorithms in Cloud
Computing,” vol. 4, no. 2, pp. 433–439, 2015.
[7] P. Singh, M. Dutta, and N. Aggarwal, “A review of task scheduling based
on meta-heuristics approach in cloud computing,” Knowl. Inf. Syst., vol.
52, no. 1, pp. 1–51, 2017.
[8] R. Eswaraprasad and L. Raja, “A review of virtual machine (VM)
resource scheduling algorithms in cloud computing environment,”
Journal of Statistics and Management Systems, vol. 20, no. 4. pp. 703–
711, 2017.
Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on March 12,2021 at 07:01:53 UTC from IEEE Xplore. Restrictions apply.
[9] S. C. Satapathy, J. K. Mandal, S. K. Udgata, and V. Bhateja, “Information
Systems Design and Intelligent Applications: Proceedings of Third
International Conference INDIA 2016, Volume 1,” Adv. Intell. Syst.
Comput., vol. 433, pp. 619–627, 2016.
[10] S. Potluri and K. S. Rao, “Quality of service based task scheduling
algorithms in cloud computing,” Int. J. Electr. Comput. Eng., vol. 7, no.
2, pp. 1088–1095, 2017.
[11] V. K. Reddy, “Articles a Survey of Various Task Scheduling,” vol. 1, no.
1, pp. 1–8, 2013.
[12] M. Sohani, S. Jain, I. Narang, K. Agarwal, and S. Arora, “A Survey of
Different Task Scheduling Algorithm in Cloud Computing,” vol. 6, no. 4,
pp. 1–7, 2017.
[13] J. Garg and G. Bhathal, “Research Paper on Genetic Based Workflow
Scheduling Algorithm in Cloud Computing,” vol. 8, no. 5, 2017.
[14] J. Ma, W. Li, T. Fu, L. Yan, and G. Hu, “A Novel Dynamic Task
Scheduling Algorithm Based on Improved Genetic Algorithm in Cloud
Computing,” vol. 348, pp. 829–835, 2016.
[15] A. Thomas, G. Krishnalal, and V. P. Jagathy Raj, “Credit based
scheduling algorithm in cloud computing environment,” Procedia
Comput. Sci., vol. 46, no. Icict 2014, pp. 913–920, 2015.
[16] T. Zhao and M. Jing, “Bandwidth-aware multi round task scheduling
algorithm for cloud computing,” J. Intell. Fuzzy Syst., vol. 31, no. 2, pp.
1053–1063, 2016.
[17] T. Kaur and I. Chana, “GreenSched: An intelligent energy aware
scheduling for deadline-and-budget constrained cloud tasks,” Simul.
Model. Pract. Theory, vol. 82, pp. 55–83, 2018.
[18] R. Van Den Bossche, K. Vanmechelen, and J. Broeckhove, “Online cost-
efficient scheduling of deadline-constrained workloads on hybrid
clouds,” Futur. Gener. Comput. Syst., vol. 29, no. 4, pp. 973–985, 2013.
[19] K. Li, G. Xu, G. Zhao, Y. Dong, and D. Wang, “Cloud task scheduling
based on load balancing ant colony optimization,” Proc. - 2011 6th Annu.
ChinaGrid Conf. ChinaGrid 2011, pp. 3–9, 2011.
[20] J. Meena, M. Kumar, and M. Vardhan, “Cost Effective Genetic Algorithm
for Workflow Scheduling in Cloud Under Deadline Constraint,” IEEE
Access, vol. 4, pp. 5065–5082, 2016.
[21] B. Wang and J. Li, “Load balancing task scheduling based on Multi-
Population Genetic Algorithm in cloud computing,” Chinese Control
Conf. CCC, vol. 2016–Augus, pp. 5261–5266, 2016.
[22] K. M. Cho, P. W. Tsai, C. W. Tsai, and C. S. Yang, “A hybrid meta-
heuristic algorithm for VM scheduling with load balancing in cloud
computing,” Neural Comput. Appl., vol. 26, no. 6, pp. 1297–1309, 2015.
[23] F. Ebadifard and S. M. Babamir, “A PSO-based task scheduling algorithm
improved using a load-balancing technique for the cloud computing
environment,” Concurr. Comput., no. October 2017, pp. 1–16, 2017.
[24] Rakotomamonjy A., “Variable Selection Using SVM-based Criteria,” J.
Mach. Learn. Res., vol. 3, pp. 1357–1370, 2003.